> ## Documentation Index
> Fetch the complete documentation index at: https://docs.atomscale.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Working in a Notebook

> Use the assistant, inspect Python cells, and access platform data

Open a notebook from the **Analysis** sidebar. If it opens on saved results, click **Edit notebook** to start editing. If another user is editing, you can review the saved results while you wait. See [Saving, Versions, and Access](https://docs.atomscale.ai/platform/reference/notebooks/saving-and-access.md).

## Use the assistant

Expand the right sidebar and select **Agent**. Send a question or instruction about the open notebook. The assistant can inspect and change its cells, load platform data, and help explain results.

Be specific about the data and purpose. For example, identify a sample by name and ask the assistant to explain how it selected and aligned the relevant data. Follow up in the same conversation to refine the analysis.

Review changes before relying on a conclusion. Check the data sources, units, intervals, assumptions, and whether the explanation is supported by the results. Ask for evidence that could contradict a proposed explanation.

The **Agent** panel is unavailable to demo accounts and while another user holds the editing session.

## Edit and run Python

The notebook editor uses marimo Python cells. Edit a cell and use its run control to execute it. Inspect cell errors and outputs before continuing. Python names shared between cells form dependencies, so changes can affect later results.

You can work directly in the cells or ask the assistant to make a change. Use **Save Changes** in the Atomscale notebook header to save a version of the notebook. Running cells is separate from saving a version.

## Access Atomscale data

The Python SDK is preinstalled and authentication is configured for the notebook. Create a client without pasting an API key:

```python theme={null}
from atomscale import Client

client = Client()
```

Use the SDK to [search data](https://docs.atomscale.ai/sdk/search-data.md), [inspect analysis results](https://docs.atomscale.ai/sdk/inspect-results.md), or work with [samples and projects](https://docs.atomscale.ai/sdk/samples-projects.md). Unlike a notebook running on your own computer, an Atomscale notebook does not need your personal API key in its source.

Check the selected data before analysis. A data item can exist before processing has completed; its [pipeline status](https://docs.atomscale.ai/platform/reference/models/data-items.md#pipeline-status) helps distinguish missing input from unavailable derived results. For data collected by a gateway, see [Finding Collected Data](https://docs.atomscale.ai/platform/reference/connecting-data/integrations.md#finding-collected-data).

## Resume a paused session

An inactive notebook session can pause. Click **Resume notebook** on the paused-session screen, or **Resume** in the banner, to reconnect. Rerun cells if their in-memory state needs rebuilding. Save important edits explicitly rather than relying on session recovery.

If an error or read-only view prevents you from continuing, see [Troubleshooting](https://docs.atomscale.ai/platform/reference/notebooks/troubleshooting.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.